Radiance Field Sampling for Unresolved Specular Object Classification
We introduce a novel technique for classifying unresolved specular objects. In remote sensing targets of interest like drones, helicopters, planes, cars, or satellites are often under-resolved. If those targets have specular surface characteristics, like metallic bodies or windows for example, there will be 'glint' instances in detection which correspond to high signal levels from the specular surface on that target. The radiance field from a specular surface will be distributed spatially in a way that depends on the shape of that surface. And, as the target, source, or imaging system moves, one can collect a time-series sample of that radiance field and use that data to classify the unresolved surface which generated that field. This allows us to extract useful target information even if we are operating in an unresolved and/or turbulent regime, since specular signals provide high-signal information about the surface compared to diffuse surfaces, especially at long ranges. By employing both active and passive illumination schemes we demonstrate this concept in the lab and at a range of 4.7 km.
